The Distinction Most Teams Get Wrong
Most non-tech companies are deploying AI in revenue workflows the same way they adopted Salesforce in 2012: buy the thing, point it at the process, hope adoption follows. The results are predictably mixed — reps using AI for the wrong tasks, managers unsure what to inspect, and enablement teams rebuilding the same playbooks every quarter.
Gong's CMO Udi Ledergor laid out a cleaner mental model in a recent conversation with Chris Walker: two buckets, hard line between them.
Bucket 1 — AI does it instead of humans: CRM logging, follow-up email drafting, extracting commitments from call transcripts. These are tasks where speed and consistency matter more than judgment. Automate them fully and stop asking reps to touch them.
Bucket 2 — AI guides, humans execute: Live customer conversations, deal qualification calls, market positioning decisions. Here the AI surfaces context, flags risk, or runs reps through simulated scenarios — but a human owns the moment.
Where the Leverage Actually Lives
The bucket framework matters because the failure mode is almost always bucket confusion. Teams automate Bucket 2 tasks (pushing AI-generated responses into live deal conversations) and keep doing Bucket 1 tasks manually (reps still updating CRM fields at end of week). Both mistakes bleed time and quality simultaneously.
Gong's own numbers from customers running AI-guided messaging are concrete: 30% higher win rates and 20% shorter sales cycles. The mechanism isn't AI replacing the rep — it's AI training the rep before the call, identifying gaps from previous recordings, running simulations, and returning them to live conversations better prepared.
That's a different ROI model than most RevOps teams are using to justify AI spend. It's not headcount replacement math. It's throughput-per-rep math.
The Actionable Takeaway
Before your next AI tool evaluation or enablement sprint, run your current workflow inventory through the two buckets. For every task on the list, answer one question: does judgment in this moment affect revenue outcome? If yes, it's Bucket 2 — AI assists, human decides. If no, it's Bucket 1 — fully automate and remove it from the rep's plate entirely.
Then look at where your actual AI budget is deployed. If most of it is in Bucket 2 automation (chatbots fielding discovery calls, AI writing cold outreach with no human editorial layer), you're optimizing the wrong bucket. The win rate data suggests the highest-leverage AI investment in revenue right now is in pre-call preparation and rep development — not in removing humans from the conversation itself.